TabPFN-4: The next step for predictive intelligence

Our next-generation tabular foundation model, TabPFN-4, has finished training. It breaks the quadratic barrier and moves the Pareto frontier on all evaluated benchmarks. Its novel sub-quadratic architecture scales to 10 million rows, while setting new Pareto frontiers in BeyondArena and TabArena.

TabPFN-4 will be available soon. Join the waitlist to get access first:

TabPFN-4 across benchmarks

Elo per benchmark (higher is better)

BenchmarkTabPFN-4TabPFN-3.5Kumo-TabularBest other baseline
BeyondArena1,5211,3701,4271,357Seldon
↳ Large tables (100k+ rows)1,5601,1641,1591,268RealMLP
TabArena1,9841,8551,9721,994LimiX-2
ScoringBench (CRPS)2,2571,9382,2282,050LimiX-2
STRABLE1,8791,6201,8281,761Chakra-Tab M
MulTaBench (text)1,8431,6371,8131,790LimiX-2
MulTaBench (image)1,8741,7621,8561,850LimiX-2
RelArena-α2,0551,914—1,908kurversc
RamanBench1,6191,4291,5561,510Chakra-Tab
TabBench-Bio1,2701,1731,2101,157Seldon-Mini
Evaluation details

Bold: best in row. TabPFN-4 with 4 estimators, one seed. Each column takes that family's best entry.

BeyondArena: TabPFN-4 sees the group id as a feature.

TabArena: the top models are within each other's error bars; LimiX-2 takes about 60× longer per prediction (see the Pareto plot).

STRABLE: every model with TF-IDF text features.

MulTaBench: text slice only; every model with frozen e5-small text embeddings.

MulTaBench image slice (40 datasets), frozen dinov3-small encoder.

RelArena-α: the TabPFN columns run the TabPFN-Rel recipe; no Kumo entry.

TabBench-Bio: TabPFN-4 is the mean of 5 seeds; no CI.

BeyondArena: TabPFN-4 leads the Pareto front

1 / 2
BeyondArena Elo (higher is better)Elo (higher is better)1,1501,2321,3141,3961,4781,5600.10.31310Other baseline (Default): 1357.90 Elo; 0.436 s / 1,000 test rowsOther baseline (Default): 1320.30 Elo; 0.242 s / 1,000 test rowsOther baseline (Default): 1315.20 Elo; 0.461 s / 1,000 test rowsOther baseline (Default): 1275.00 Elo; 9.163 s / 1,000 test rowsOther baseline (Default): 1242.50 Elo; 0.830 s / 1,000 test rowsOther baseline (Default): 1229.40 Elo; 3.310 s / 1,000 test rowsOther baseline (Tuned + Ens.): 1224.20 Elo; 14.402 s / 1,000 test rowsOther baseline (Default): 1173.70 Elo; 2.393 s / 1,000 test rowsOther baseline (Tuned + Ens.): 1171.60 Elo; 6.092 s / 1,000 test rowsOther baseline (Tuned + Ens.): 1170.10 Elo; 1.103 s / 1,000 test rowsOther baseline (Default): 1168.30 Elo; 2.419 s / 1,000 test rowsOther baseline (Tuned): 1166.70 Elo; 1.699 s / 1,000 test rowsOther baseline (Tuned): 1156.80 Elo; 0.208 s / 1,000 test rowsOther baseline (Default): 1154.20 Elo; 0.296 s / 1,000 test rows1 est.2 est.4 est.Seldon-MiniKumo-Tabular SLimiX-2TabPFN-3.5Kumo-Tabular MKumo-Tabular LInference seconds / 1,000 test rows (log scale)
TabPFN-4TabPFN-3.5Named baselinesOther baselines
Evaluation details

Baselines as of November 17, 2026. TabPFN-4 timed on an RTX PRO 6000 following the official benchmark protocol. Median inference time per 1,000 test rows. Dashed line: Pareto front.

BeyondArena Improvability (%) (lower is better)Improvability (%) (lower is better)68.210.412.614.8170.10.31310Other baseline (Default): 9.89%; 0.436 s / 1,000 test rowsOther baseline (Default): 12.58%; 0.242 s / 1,000 test rowsOther baseline (Default): 12.74%; 0.461 s / 1,000 test rowsOther baseline (Default): 11.32%; 9.163 s / 1,000 test rowsOther baseline (Default): 14.75%; 0.830 s / 1,000 test rowsOther baseline (Default): 15.11%; 3.310 s / 1,000 test rowsOther baseline (Tuned + Ens.): 16.03%; 14.402 s / 1,000 test rowsOther baseline (Default): 16.48%; 2.419 s / 1,000 test rows1 est.2 est.4 est.Kumo-Tabular LKumo-Tabular MLimiX-2TabPFN-3.5Kumo-Tabular SSeldon-MiniInference seconds / 1,000 test rows (log scale)
TabPFN-4TabPFN-3.5Named baselinesOther baselines
Evaluation details

Baselines as of November 17, 2026. TabPFN-4 timed on an RTX PRO 6000 following the official benchmark protocol. Median inference time per 1,000 test rows. Dashed line: Pareto front.

Scaling to 10 million rows

Extrapolated: 3.5 ≈65 h, Kumo ≈127 h

1 / 2

Extrapolated: 3.5 ≈65 h, Kumo ≈127 h

Prediction time up to 10 million training rows: quadratic extrapolationPrediction time (hours)02040608010012014002.5M5M7.5M10MTabPFN-4: 38 min (measured)TabPFN-3.5: ≈65 hKumo-Tabular L: ≈127 hTraining rows
TabPFN-4TabPFN-3.5Kumo-Tabular L

dashed: quadratic fit, extrapolated

Evaluation details

Synthetic data, 500 columns, 1,000 test rows, one GPU.

One estimator timed, scaled by the default count.

Measured: 38 min at 10M vs 48 min at 1M

Prediction time up to 10 million training rows: measured timingsPrediction time (minutes)0102030405002.5M5M7.5M10MTabPFN-4: 38 minTabPFN-3.5: 48 min at 1M rowsKumo-Tabular L: 13 min at 262k rowsOut of GPU memory above 262k rowsTraining rows
TabPFN-4TabPFN-3.5Kumo-Tabular L
Evaluation details

Synthetic data, 500 columns, 1,000 test rows, one GPU.

One estimator timed, scaled by the default count.

Previous tabular foundation models scale quadratically with the number of rows, so large tables quickly become impractical. Extrapolated to 10 million training rows, TabPFN-3.5 would take about 65 hours and Kumo-Tabular L about 127 hours at their default estimator counts. TabPFN-4’s subquadratic architecture runs the same 10 million × 500 table in 38 minutes on a single RTX PRO 6000, roughly 100× faster than TabPFN-3.5 and 200× faster than Kumo-Tabular L. This enables fast experimentation at larger scales than any previous tabular foundation model.

Biggest gains on large, real-world data

TabPFN-4 improvements are particularly large in data regimes that are common among high-impact real-world use cases: large, temporal, and grouped data.

Until now, tuned models such as XGBoost and RealMLP, or Kumo-Tabular, were the strongest baselines in these regimes. On BeyondArena, where TabPFN-4 receives the group ID as a feature, it leads the best of them by 292 Elo on tables above 100,000 rows, 229 on temporal splits and 145 on grouped splits, compared with 42 on random splits.

Temporal and grouped data break the assumption that rows are independent draws from one distribution: the future differs from the past, and rows from the same customer or machine look alike. Classical models only handle this when someone engineers it into features. TabPFN-4’s lead is largest exactly here, which suggests a foundation model can learn to handle such structure on its own.

BeyondArena: large tables

1 / 4

BeyondArena: large tables

2,0001,8001,6001,4001,2001,000800
1,168
XGBoost
1,268
RealMLP
912
TabPFN-3
1,164
TabPFN-3.5
1,560
TabPFN-4
Elo (95% confidence intervals); bars start at 1,000
Evaluation details

BeyondArena, baseline set 1117 (official); TabPFN-4 sees the group id as a feature. XGBoost (default) = 1,000. 95% bootstrap CIs. One seed.

BeyondArena: temporal split

1,8001,6001,4001,2001,000800
1,126
XGBoost
1,173
Kumo-L
1,023
TabPFN-3
1,172
TabPFN-3.5
1,403
TabPFN-4
Elo (95% confidence intervals); bars start at 1,000
Evaluation details

BeyondArena, baseline set 1117 (official); TabPFN-4 sees the group id as a feature. XGBoost (default) = 1,000. 95% bootstrap CIs. One seed.

BeyondArena: grouped split

1,6001,4001,2001,000800
1,095
XGBoost
1,246
RealMLP
1,048
TabPFN-3
1,156
TabPFN-3.5
1,391
TabPFN-4
Elo (95% confidence intervals); bars start at 1,000
Evaluation details

BeyondArena, baseline set 1117 (official); TabPFN-4 sees the group id as a feature. XGBoost (default) = 1,000. 95% bootstrap CIs. One seed.

BeyondArena: random split

1,8001,6001,4001,2001,000
1,122
XGBoost
1,574
Kumo-L
1,333
TabPFN-3
1,496
TabPFN-3.5
1,616
TabPFN-4
Elo (95% confidence intervals); bars start at 1,000
Evaluation details

BeyondArena, baseline set 1117 (official); TabPFN-4 sees the group id as a feature. XGBoost (default) = 1,000. 95% bootstrap CIs. One seed.

The strength of TabPFN-4 in these data regimes directly leads to large improvements on benchmarks that make use of such messy, complex, real-world data. Just by switching out the model in the TabPFN-Rel harness, we see an uplift of over 100 Elo points on relational data (RelArena-alpha). On high-dimensional biomedical data (TabBench-Bio), TabPFN-4 outperforms the next-best baseline by over 50 Elo points on average.

RelArena-α: TabPFN-4 against the top relational models and systems

1 / 2

RelArena-α: TabPFN-4 against the top relational models and systems

TabPFN-4
2,055
TabPFN-Rel API (3.5)
1,914
kurversc
1,908
TabPFN-Rel local (3.5)
1,878
relflow
1,855
RT-J
1,839
RT-PluRel
1,784
1,5001,7001,9002,1002,300
Elo (95% confidence intervals); axis starts at 1,500
Evaluation details

RelArena-α public leaderboard, models and systems; newest entry per family, plus TabPFN-Rel local. Constant baseline = 1,000.

95% bootstrap CIs over tasks. One seed.

TabBench-Bio: mean Elo by task, sample budget and feature cap

TabBench-Bio mean Elo by task, sample budget and feature cap10001100120013001400AllClassificationRegression20 samples50 samples100 samples200 samples500 samples1,000 samplesAll samples2,000 features10,000 features25,000 featuresAll features
TabPFN-4TabPFN-3.5Kumo-Tabular LMLPCatBoost
Evaluation details

Mean Elo over the benchmark's 28 settings. TabPFN-4 was evaluated by us with the benchmark's pipeline, all other models as published.

Tables with text and images

We evaluate TabPFN-4 on two multimodal tabular benchmarks, STRABLE and MulTaBench. Every model gets the same encoder: TF-IDF for STRABLE, e5-small-v2 for MulTaBench Text, and DINOv3-small for MulTaBench Image. TabPFN-4 ranks first on all of them.

On the text benchmarks we also show TabPFN-4 Plus, the default in our API, whose built-in text handling pushes the lead further.

STRABLE

1 / 3

STRABLE

TabPFN-4 Plus
2,010
TabPFN-4 (TF-IDF)
1,879
Kumo-Tabular (TF-IDF)
1,828
TabPFN-3.5 (TF-IDF)
1,620
1,5001,6621,8251,9882,150
Elo (95% confidence intervals); axis starts at 1,500

MulTaBench (text)

TabPFN-4 Plus
1,917
TabPFN-4
1,843
Kumo-Tabular (Frozen-S)
1,813
TabPFN-3.5 (Frozen-S)
1,637
1,5001,6621,8251,9882,150
Elo (95% confidence intervals); axis starts at 1,500

MulTaBench (images)

TabPFN-4
1,874
Kumo-Tabular (Frozen-S)
1,856
TabPFN-3.5 (Frozen-S)
1,762
1,5001,6621,8251,9882,150
Elo (95% confidence intervals); axis starts at 1,500

TabPFN-4 on Kaggle

Tabular competitions have always been among the most contested on Kaggle. Winning meant months of feature engineering, hyperparameter tuning, and deep ensembles, built against a private test set nobody got to see. Tabular foundation models have been improving rapidly over the past few months and have reached a point where they would have matched or beaten the winners of several of those competitions with a single fit.

We made late submissions with TabPFN-4 to 38 closed competitions: 25 Playground Series episodes and 13 older prize-money competitions. One forward pass per competition with default settings, no tuning, scored on the private leaderboard.

Prize-money competitions (13 datasets)

1 / 2

Prize-money competitions (13 datasets)

TabPFN-4
71311
TabPFN-3.5
4135
Kumo-Tabular
51322
1st-place scoreGold zoneSilver zoneBronze zoneBelow bronze
Evaluation details

Private-leaderboard rank of one late submission per competition: one fit on the raw training data, default settings, no tuning, one run.

Zones follow Kaggle's medal cut-offs for the number of teams; late submissions earn no medals. Rank = 1 + teams with a strictly better score.

Playground competitions (25 datasets)

TabPFN-4
723112
TabPFN-3.5
253411
Kumo-Tabular
221416
1st-place scoreGold zoneSilver zoneBronze zoneBelow bronze
Evaluation details

Private-leaderboard rank of one late submission per competition: one fit on the raw training data, default settings, no tuning, one run.

Zones follow Kaggle's medal cut-offs for the number of teams; late submissions earn no medals. Rank = 1 + teams with a strictly better score.

TabPFN-4 would have tied or beaten the original winner on 14 of the 38 and finished in the gold zone on 8 of the 13 prize-money competitions. On prize-money competitions, that is 7 wins, two more than Kumo-Tabular and three more than TabPFN-3.5. In the 25 Playground Series competitions, it would have taken first place on the private leaderboard in 7, ahead of every human team; Kumo-Tabular and TabPFN-3.5 would have won 2 each. Tabular foundation models have become the de facto starting model for Kagglers on tabular problems, and TabPFN-4 is now the strongest of them.

What's Next

We are working with customers to apply TabPFN-2.5 Scaling Mode to their largest datasets.